State voter turnout will likely reach quarter century high for mid-term elections this Fall

EXCLUSIVE MPNblog.com ANALYSIS

Electoral winners in Missouri are always at the mercy of voter turnout. That is, if one party's voters go to the polls to cast votes while the other party's members stay home because of weather, apathy, or some other factor, the election may swing unpredictably for or against a particular candidate.

So what should we expect voter turnout to look like in Missouri for 2010?

Using widely accepted statistical techniques, MPNblog.com developed a model to project statewide voter turnout in 2010, and if our results fall within a 95 percent confidence interval, Missouri will see the greatest mid-term election turnout in 30 years.

Voter Turnout Model
Our OLS (ordinary least squares regression) model accounts for just under 100 percent of the variation in voter turnout using three statistically significant variables -- senate election, presidential/gubernatorial election, and the statewide unemployment rate -- while controlling for time and the state's voting age population.

The key drivers of voter turnout, not unexpectedly, are what races are up for election in a given political cycle. A dummy variable for presidential and gubernatorial elections captures the largest impacts on voter turnout in our model and is highly statistically significant (p<0.001). In years of presidential and gubernatorial elections, voter turnout is projected to increase by 705,831 (+/- 64,000), ceteris paribus. While the impact of senate elections is less statistically significant, these years can drive turnout up by 141,151 (+/- 72,560).

The seasonally adjusted Nov. unemployment rate was incorporated to capture an element of voter dissatisfaction. Incidentally a higher unemployment rate would also reduce one of the common barriers to voting -- getting away from work on election day. This variable, while statistically significant, has only a marginal impact on turnout. For a one point increase in the state's seasonally adjusted unemployment rate, voter turnout can increase 69,170 (+/- 33,606). We built into the model a control for voting age population, based largely on imputations from the Census Bureau's annual data that estimates total population as of July 1. The imputation adjusted the annual population projections to estimate the total population 18 or older in Nov., assuming population growth occurs linearly over the course of a year.  

Limitations of our Model
Before we reveal the projection, in the interest full disclosure we note several limitations of this aggregate model's application.

First, we note that House races were not built into this model. Since all House seats are up for re-election every year, we generally do not expect that they drive voter turnout, but rather are present in the baseline. That said, this year may present particular anomalies due to the targeting of incumbents like U.S. Rep. Ike Skelton (D) in Missouri's Fourth Congressional District. Such anomalies will create some level of variation in total participation as both anti-Skelton voters and pro-Skelton voters may be guaranteed to come to the polls, even though in past years they may have been more apathetic in mid-term election cycles.

We also note that this model has not yet incorporated county-level variation into the turnout projection. As we noted for the 2008 primary and general elections, turnout rates vary greatly across the state based on a series of socioeconomic, geographic, and even meteorological factors. For instance, it's feasible that in the absence of a high-profile African-American Presidential candidate, African-Americans in St. Louis City will be less likely to vote in the 2010 general election thus decreasing overall turnout.

Third, although the variation accounted for in this model is fairly high, the omission of House-level data may be problematic looking at historical results. In 1994, when Republicans captured the House of Representatives in a mid-term election, our model projected  41 percent turnout while actual turnout was 5 points higher, at 46 percent. This is important because there's no simple, empirical method to include the impacts of issues such as health care reform and the Tea Party movement, which may also impact turnout rates.

Fourth, we make only a rough attempt to project the state unemployment rate. Based on current trends, general economic forecasts from Creighton University, and national projections from the White House Office of Management and Budget, we assume unemployment rates will decline marginally in coming months.

Finally, our model is constructed with a relatively small subset of data which may impact the validity of the results, but we have examined the model projections and compared to the actual historical turnout rates, and find the deviations to be small enough to suggest this model is fairly good at explaining turnout (also evidenced by the r-squared of 0.96).

MPN's 2010 Statewide Voter Turnout Projection
So what do we expect voter turnout to look like in 2010? We estimate that turnout will be about 55.6 percent of the voting age population, or 2.5 million voters.

This assumes that by Nov. 2010, the state's unemployment rate declines to 9.0 percent. If that rate doesn't change, and remains at the Feb. 2010 level of 9.4 percent, voter turnout could be as high as 56.2 percent.

With high unemployment this year and the U.S. Senate contest driving voters to the polls, we project turnout will be a full 7 points above the best turnout in a mid-term elections since the 1980s. In fact the lowest turnout in those races -- the 1990 mid-term when unemployment was just 6.5 percent -- saw turnout of just 35 percent, so this contest could be a full 21 points higher.

Our projection would result in about 417,000 fewer votes cast than in the 2008 presidential contest in Missouri, but 380,000 more than in 2006 when Democrats nationally recaptured the House from Republicans.

The thing to keep in mind, though, is that while this turnout rate will likely be considerably high for a mid-term election, it's still low compared to a presidential contest. Our mean projection would be greater than just one presidential contest in the last 30 years, the 1996 re-election of Bill Clinton where Missouri turnout was about 54 percent.

The results could have broad implications for candidates not only engaged in national contests, but also for those involved in more localized races like the state Senate and House contests. Our earlier projections for House races and Senate races did not incorporate any analysis of likely voter turnout, but the next iteration to be released in the near future will build in some assumptions based on this analysis.

We won't get into which party benefits more from this higher turnout rate in this post, so look forward to that  analysis in the future.

Related Posts by Category: